US6185299B1ExpiredUtility

Adaptive echo cancellation device in a voice communication system

58
Assignee: IBMPriority: Oct 31, 1997Filed: Jun 22, 1998Granted: Feb 6, 2001
Est. expiryOct 31, 2017(expired)· nominal 20-yr term from priority
H04M 9/082
58
PatentIndex Score
28
Cited by
5
References
12
Claims

Abstract

A method and device for adaptive echo cancellation in a voice communication system wherein a reference signal x is sent over the system from a transmitting device and comprising an adaptive filter for removing any echo signal from a primary signal d received in answer to reference signal x, and said adaptive filter having a plurality of coefficients the value of which at each time n being computed from the value of the same coefficient at time n−1 modified by a normalization factor depending upon the energy of said reference signal and provided by a normalization factor controlling unit. The normalization factor is the maximum value between a first value representing the energy of reference signal x at time n and a second value depending on a value of the normalization factor previously stored during a preceding predetermined interval time multiplied by an attenuation factor less than 1.

Claims

exact text as granted — not AI-modified
What is claimed:  
     
       1. A method for adaptive echo cancellation in a voice communication system comprising the steps of: 
       transmitting a reference signal x over the system from a transmitting device;  
       removing any echo signal from a primary signal d received by said transmitting device, said echo signal being sent back by said voice communication system in answer to said reference signal x, using an adaptive filter having a plurality of coefficients the value of which at each time n being computed from the value of the same coefficient at time n−1 modified by a normalization factor depending upon the energy of said reference signal d in that said normalization factor is the maximum value between a first value representing the energy of said reference signal at time n and a second value depending on a value of said normalization factor previously stored during a preceding predetermined interval time multiplied by an attenuation factor less than 1.  
     
     
       2. The method according to claim  1 , wherein the values of the adaptive filter coefficients are given by the formulae          F        (     n   +   1     )       =       F        (   n   )       +       μ     a   +     A        (   n   )           ·     X        (   n   )       ·     e        (   n   )                           
       and said normalization factor has the following value: A(n)=max[η·Amax, ∥X(n)∥ 2 ] in which μ is an adaptation speed factor, a is a small positive stabilization constant, η is an attenuation factor less than 1, Amax is the maximum value of said normalization factor stored during a previous predetermined interval time, X(n)=[x(n), x(n−1), . . . ,x(n−L+1)] with L being the number of filter taps in the adaptive filter, and e(n)=d(n)−F(n)·X(n) where d(n) is the value of primary signal at time n. 
     
     
       3. The method according to claim  2 , wherein the value of Amax is the value A(n−1) of said normalization factor at time n−1. 
     
     
       4. The method according to claim  1 ,  2  or  3 , wherein η is approximately=0.9999. 
     
     
       5. A device for adaptive echo cancellation in a voice communication system comprising: 
       a circuit for transmitting a reference signal x over said communication system and receiving a primary signal d from said voice communication system, the primary signal including an echo signal from the voice communication system in response to the transmission of the reference signal x;  
       an adaptive filter for removing any echo signal sent back by said voice communication system in answer to said reference signal and a normalization factor controlling unit for determining the value at time n of each coefficient of said adaptive filter based upon the value of the same coefficient at time n−1 modified by a normalization factor A(n) depending on the energy of said reference signal;  
       the normalization factor controlling unit having a means for selecting said normalization factor as being the maximum value between a first value representing the energy of said reference signal at time n and a second value depending on a value of said normalization factor previously stored during a preceding predetermined interval time multiplied by an attenuation factor.  
     
     
       6. The device according to claim  5 , wherein said normalization factor controlling unit further comprises: 
       a first means for determining the maximum between ∥X(n)∥ 2  and η·Amax wherein X(n)=[x(n),x(n−1), . . . ,x(n−L+1)] with L being the number of filter taps in the adaptive filter, η is an attenuation factor less than 1, Amax is the maximum value of said normalization factor stored during a previous predetermined interval time, and  
       second means for determining        μ     a   +     A        (   n   )                         
       where a is a small positive stabilization constant, μ is an adaptation speed factor, and A(n) is said normalization factor at time n; 
       a multiplier to determine a          μ     a   +     A        (   n   )           ·     X        (   n   )       ·     e        (   n   )                       
       wherein e(n)=d(n)−F(n)·X(n) with d(n) being the value of said primary signal at time n and F(n) are the values of the adaptive filter coefficients; 
       a summing circuit for determining the values of said adaptive filter coefficients in accordance with the formulae          F        (     n   +   1     )       =       F        (   n   )       +       μ     a   +     A        (   n   )           ·     X        (   n   )       ·       e        (   n   )       .                         
     
     
       7. The device according to claim  6 , wherein the value of Amax is the value A(n−1) of said normalization factor at time n−1. 
     
     
       8. The device according to claim  6 , wherein η is approximately 0.9999. 
     
     
       9. A method for adaptive echo cancellation in a communication system comprising the steps of: 
       transmitting a reference signal x to the system;  
       receiving a primary signal d from the system, the primary signal including an echo signal from the communication system in response to the transmission of the reference signal x;  
       removing the echo signal from the primary signal d using an adaptive filter having a plurality of coefficients the value of which at each time n being computed from the value of the same coefficient at time n−1 modified by a normalization factor depending upon the energy of said reference signal d in that said normalization factor is the maximum value between a first value representing the energy of said reference signal at time n and a second value depending on a value of said normalization factor previously stored during a preceding predetermined interval time multiplied by an attenuation factor less than 1.  
     
     
       10. The method according to claim  9 , wherein the values of the adaptive filter coefficients are determined by          F        (     n   +   1     )       =       F        (   n   )       +       μ     a   +     A        (   n   )           ·     X        (   n   )       ·     e        (   n   )                           
       and said normalization factor is 
       A(n)=max[η·Amax, ∥X(n)∥ 2 ] in which μ is an adaptation speed factor, a is a small positive stabilization constant, η is an attenuation factor less than 1, Amax is the maximum value of said normalization factor stored during a previous predetermined interval time, X(n)=[x(n), x(n−1), . . . ,x(n−L+1)] with L being the number of filter taps in the adaptive filter, and e(n)=d(n)−F(n)·X(n) where d(n) is the value of primary signal at time n.  
     
     
       11. The method according to claim  10 , wherein the value of Amax is the value A(n−1) of said normalization factor at time n−1. 
     
     
       12. The method according to claim  9 ,  10 , or  11 , wherein η is approximately=0.9999.

Cited by (0)

No later patents cite this yet.

References (0)

No backward citations on record.